University of Wollongong

Research Online
Faculty of Engineering and Information Sciences Papers

Faculty of Engineering and Information Sciences

2000

Developments in blast furnace process control at

Port Kembla based on process fundamentals
Robert Nightingale
BHP Steel, robertn@uow.edu.au

Rian Dippenaar
University of Wollongong, rian@uow.edu.au

Wei-Kao Lu
McMaster University

Publication Details
Nightingale, R. J., Dippenaar, R. J. & Lu, W. (2000). Developments in blast furnace process control at Port Kembla based on process
fundamentals. Metallurgical and Materials Transactions B, 31 (5), 993-1003.

Research Online is the open access institutional repository for the

University of Wollongong. For further information contact the UOW
Library: research-pubs@uow.edu.au

Developments in blast furnace process control at Port Kembla based on

process fundamentals
Abstract

This article is based on a presentation made in the "Geoffrey Belton Memorial Symposium," held in January
2000, in Sydney, Australia, under the joint sponsorship of ISS and TMS.
Keywords

port, control, developments, process, blast, kembla, fundamentals, furnace

Disciplines

Engineering | Science and Technology Studies

Publication Details

Nightingale, R. J., Dippenaar, R. J. & Lu, W. (2000). Developments in blast furnace process control at Port
Kembla based on process fundamentals. Metallurgical and Materials Transactions B, 31 (5), 993-1003.

This journal article is available at Research Online: http://ro.uow.edu.au/eispapers/2663

Developments in Blast Furnace Process Control at Port

Kembla Based on Process Fundamentals
ROBERT J. NIGHTINGALE, RIAN J. DIPPENAAR, and WEI-KAO LU

I. THE BLAST FURNACE PROCESS

THE iron blast furnace is in principle a countercurrent
gas/solid heat exchanger from tuyere raceway to stockline
and a countercurrent oxygen exchanger from fusion zone to
stockline.[1] Solid raw materials consisting of iron ore, sinter,
coke, and fluxes are charged into the top of the furnace,
while air, and sometimes hydrocarbons and oxygen, is
blasted through tuyeres near the bottom of the furnace. The
retention time of the ore may be as long as 8 hours, while
that of the gas is a few seconds. However, the residence
time of the coke in the hearth is longer, varying from 1 to
4 or more weeks.[2] The liquid hot metal and slag products
are tapped at regular intervals through several tapholes near
the bottom of the furnace. The slag is separated from the
metal and the liquid metal is transported to the steel plant.
In an attempt to gain a better understanding of this complex process in which direct experimental measurement is
exceedingly difficult, experimental as well as operating furnaces have been quenched and dissected. Earlier studies by
Bosley et al.[3] and Muravev et al.[4] were followed by very
detailed and comprehensive investigations in Japan. The
results of these studies have been fully documented, summarized, and thoroughly analyzed by a committee of the Iron
and Steel Institute of Japan.[5] Largely based on the information gained from these dissections, it has been possible to
divide the modern blast furnace for the sake of convenience
and further discussion into five different zones.[1,6]
(1) lumpy zone (upper stack and cyclic reduction zone),
(2) cohesive zone (softening and melting zone),
(3) active coke zone,
(4) HearthDeadman, and
(5) tuyere raceways.
The verification of the existence of a cohesive zone
through the dissection of quenched furnaces gave an enormous boost to blast furnace operational improvements. In
this zone, the first slag forms, ferrous materials soften, and
melting begins. In the cyclic reduction zone, carbon monoxide reacts with wustite to produce solid iron and carbon
dioxide. The carbon dioxide, in turn, reacts with coke to
regenerate carbon monoxide and this cycle prevails at temperatures above 1000 8C. Excess carbon monoxide reduces
hematite and magnetite to wustite in the upper stack.
ROBERT J. NIGHTINGALE, Superintendent of Ironmaking Development, is with BHP Steel, Port Kembla, NSW2505, Australia. RIAN J.
DIPPENAAR, Director, is with the BHP Institute for Steel Processing and
Products, University of Wollongong, Wollongong, NSW2522, Australia.
WEI-KAO LU, Professor Emeritus, is with the Department of Materials
Science, McMaster University, Hamilton, ON, Canada L8S 4L7.
This article is based on a presentation made in the Geoffrey Belton
Memorial Symposium, held in January 2000, in Sydney, Australia, under
the joint sponsorship of ISS and TMS.
METALLURGICAL AND MATERIALS TRANSACTIONS B

In the hearth, refining of the liquid metal and final reduction and separation of the liquid slag and metal occur as
well as the final carburization of the iron. In the raceway
region, oxygen in the incoming hot-blast air reacts with
carbon to produce carbon dioxide as well as heat and, subsequently, carbon monoxide. Intensive heating and reduction
take place in the active coke zone as well as in the zone of
the coke percolators. The role that this near stationary zone,
which is often referred to as the deadman, plays in effective
blast furnace operations has been clearly highlighted and its
significance in modern, large furnaces has been verified.[5]
The dissection of blast furnaces clarified in much more
detail than before the importance of the cohesive zone, the
nature of the flow of gas through the coke slits in the cohesive
zone, and the nature of the chemical reactions in the furnace.
It further drastically altered the very concept of the process.
These advances generated, in turn, a demand for improved
blast furnace instrumentation[5] and the concomitant process
control. Poos[7] reiterated these sentiments by pointing out
that blast furnace control panels of the 1950s were so rudimentary that they informed staff of only a few variables
such as stockline movement, blast pressure, and temperature.
The furnaces of that period had to be relined frequently; the
availability was below 75 pct of calender time. That was 50
years ago. However, in the last 25 years, tremendous strides
have been made in the optimization of the process. In general, the productivity of furnaces has increased, not only due
to careful control but also to increased blast temperature,
oxygen enrichment, increased top pressure, improved quality
control of raw materials, size, and size distribution control
of raw material and burden distribution control.[1,6]
The modern blast furnace is characterized by a large
hearth diameter, hindering the flow of gas as well as liquid
metal and slag through the deadman and hearth. The raceway penetrates only to a depth of 2 m into the furnace so
that gas flow is more directed to the periphery as the hearth
diameter increases and, hence, the percentage active area
diminishes with an increase in hearth diameter. In a blast
furnace with a hearth diameter of 4 m, the active ring
covers the full area at the level of the raceway. However,
the active area reduces to less than 50 pct if the hearth
diameter increases to 14 m.[8] Because of this variance in
the radial direction, the ability to manage and control gas
flow and gas distribution, therefore, is relatively more
important in big furnaces. The large inactive zone in the
furnace also impacts on the ability to drain the metal and
slag, especially with regard to maintaining core temperature
and permeability.
II. ADVANCES IN CONTROL
The development of reliable, rapid, and affordable techniques to sample and analyze the liquid product streams
VOLUME 31B, OCTOBER 2000993

enabled the first significant advances in process control.

Since these streams have long residence times and were still
infrequently sampled, strategies based on simple feedback
were quite limited. Control abilities were next improved
with the advent of continuous top gas analysis technology.
The far shorter residence time of the gas stream gave a
valuable improvement in both the response time and the
ability to close mass and energy balances on the process.
The conceptual advances derived from quenched furnace
investigations have since been enhanced by the development
of an array of sophisticated probes. Fixed probes monitor
the radial temperature distribution of the top gas stream.
Profile meters measure the placement of raw materials at
the stockline, and retractable intrusive probes routinely measure the radial variation in gas temperature and composition
within the lumpy zone.
Today, comprehensive computer based heat and mass balance modeling techniques are used to control the state of
the furnace in real time. Operators are also supported by
online models describing burden distribution and cohesive
zone formation.[9]
III. EMERGENT CHALLENGES
Business realities continue to demand that the blast furnace process be operated at increasing standards of process
safety and stability, hot metal quality, and frequently
increased productivity.
In order to achieve this, it is becoming increasingly
important to control the whole process. This necessarily
includes the region below the cohesive zone where the presence of liquid phases adds greater complexity. The physical
conditions of operating pressure and temperature in this
region make the online use of retractable probes prohibitively difficult and expensive for general application. The
notable exception is at NSC Oita,[10] where a massive watercooled probe is available. This device represents a large
capital outlay and its infrequent mode of use limits data
analysis. This is especially true in times of difficult operations when it is not used for safety reasons. These are, of
course, the most important conditions to investigate.
Probings of the coke bed through the tuyeres at shutdown
are widely practiced. Unfortunately, online probing at sensible cost and frequency is impractical. At Port Kembla and
elsewhere,[11,12] these offline efforts confirm the aliveness
and critical importance of the deadman through observed,
and all too often, inadequately explained variation in condition. These probes provide a good snapshot but they are
costly, and sample processing is so time consuming that the
acquired data cannot be used for any real time process
control.
Undoubtably, the greatest and largely unaddressed challenges are those associated with the understanding and management of the condition and flow of the coke bed in the
furnace deadman and hearth. The fundamental properties of
this process stream are critical, but the literature contains
little or no discussion on measured change in these properties
over time nor on modeling of responses to fluctuation in
voidage or cleanliness (fines contamination) of this coke bed.
To allow constructive interpretation of furnace performance in response to variations in the coke bed, a continuous
supply of data relating to the permeability of the coke bed
(especially to liquid streams) is required. Further, methods
994VOLUME 31B, OCTOBER 2000

for interpreting the data and for implementing timely control

practices need to be developed. These practice initiatives
may involve modifications to raw material properties or to
furnace operating indexes.

IV. INTERACTIONS BETWEEN LIQUIDS AND

THE SLOW MOVING COKE BED
A. The Coke Bed
Following discharge from the lower surface of the cohesive zone, metal and slag droplets drain through a packed
coke bed before collecting in the furnace hearth. Coke bed
contact continues for finite but differing times for each collected phase before removal from the furnace hearth.
The coke lumps in this part of the furnace are the survivors
of the physical and chemical rigours of prior handling and
passage through the upper parts of the furnace. As such, the
coke bed is expected to dynamically reflect changes in input
coke quality. These are usually derived from variations in
the coal blend and its preparation and to changes in coking
battery operating practice. Adjustments to blast furnace feed
conditions (e.g., alkali loadings) and operating intensity may
also influence coke degradation. The level of process stability required for viable contemporary operation dictates that
changes such as these occur infrequently.
As a result, changes in the coke bed of the blast furnace
deadman and hearth generally occur over days and weeks.
This behavior is reinforced by the solid flow dynamics of
the process itself and this has been confirmed by quenched
furnace investigations,[13,14,15] use of radioactive isotopes,[16]
and laboratory modeling[17] in addition to widespread anecdotal evidence.
The presence of fine solid particles impairs the permeability of a fixed or slow moving bed by increasing the fluidsolid contact area and friction forces. In the case of the blast
furnace deadman, three sources of fine material may broadly
be identified. The first is fines carried down from above
where they may have been introduced with the charge. They
may also be the result of volume and surface breakage
responses to the forces of the lumpy and cohesive zones. In
the case of coke, there may also be the abraded products of
surface weakening due to the carbon solution loss reaction.
The second potential source of fines is the unfluxed oxide
particles. Species such as SiO2, Al2O3, TiO2, CaO, and MgO
or particles extremely rich in these species are generally
solid at deadman temperatures. Solid particles may be deposited from dripping slags, which still contain a dispersed solid
phase at discharge. Additionally, oxide particles may be
precipitated by dripping slags where FeO serves as the flux.
In such cases, subsequent removal of the FeO on reaction
with contacted coke particles raises the liquidus temperature
of the remaining slag until solidification occurs. This behavior is possible with both highly acidic and highly basic slags,
but is more likely to be a problem with highly acidic SiO2
or TiO2 based slags due to their much lower reducibility.
Highly basic slags are likely to drip with very low residual
FeO contents. Solid oxide particles can be dynamically
removed by dissolution into the compositionally suitable
contacting slag droplets. This mechanism also serves to dissolve solid ash from the surface of coke particles.
The third source of solids is the coke in raceways. These
METALLURGICAL AND MATERIALS TRANSACTIONS B

carbonaceous fines may be coke debris or incompletely combusted char from pulverized coal injection. Raceway gases
carry these particles into the deadman and the active coke
zone. The behavior of these particles may vary widely
depending on their size, the chemical reactivity of the carbon,
and the nature of associated oxide (ash) phases. Because
gas velocities in the deadman are lower than in the active
coke zone (especially near the furnace center) the deposition
of larger particles, often those derived from coke, is a particular threat in this region. The deposition of graphite (kish)
from tiny droplets of iron expelled from the furnace raceway
is also possible since sites remote from the raceway are at
considerably lower temperature. Graphite is variably
observed in shutdown tuyere coke probings, and this may
reflect the fact that the probability of its deposition is greater
in an operation with unstable heat balance.

B. The Liquids
The blast furnace ferrous burden is most commonly a
mixture of up to five discrete materials. Of these, the most
commonly used is sinter, which is itself an inhomogeneous
agglomerate of diverse fine ores and fluxes. While some
mixing of these materials always occurs (and is often promoted), individual solid particles generally give rise to separate liquid dripping products. Therefore, the furnace
deadman represents a highly heterogeneous reaction environment as droplets of metal and slag move through the
coke bed and through gradients of temperature and reduction
potential. While laboratory experiments have been devised
to study the meltdown behavior of individual and mixed
ferrous materials, very little is known about the behavior or
mixing of molten products within the deadman.
The collected pools of metal and slag in the furnace hearth
represent a much more homogeneous condition, but still the
opportunity for heterogeneous reaction exists as the metal
droplets pass though the slag layer, at the interfacial surface
and at contact surfaces with the coke.

C. The Interactions
When the coke bed in the blast furnace deadman has high
permeability, the dripping metal and slag products move
rapidly into the hearth under the influence of gravity. The
individual transit time may, of course, vary with viscosity
in the case of slag droplets.
When liquids pass through the coke bed in the deadman/
hearth, there are three types of reaction: i.e. coke/metal,
coke/slag, and metal/slag reactions. In these heterogeneous
reactions, the kinetics of reaction depends on the area of
contact and the time of contact of reacting phases. Proper
treatment of these heterogeneous reactions may be found
elsewhere.[18] For the present work, it is sufficient to consider
the following reactions.
Carburization:
C (coke) 1 Fe(1) hot metal
When this reaction is completed, hot metal becomes
saturated with carbon, the value being determined by
temperature.
METALLURGICAL AND MATERIALS TRANSACTIONS B

Reduction of oxides in slag:

C (coke) 1 FeO (slag) [Fe] (hot metal) 1 CO (gas)
and
C (coke) 1 MnO (slag) [Mn] (hot metal) 1 CO (gas)
and
C (coke) 1

1
1
TiOx (slag) [Ti] (hot metal) 1 CO (gas)
x
x

It is well known that these reactions do not reach equilibrium states in the blast furnace hearth.
The extent of completion of these reactions depends on
many factors, mainly properties of the coke bed and reaction
conditions such as temperature, contact time, and chemical
composition of each phase.
Slag/metal reactions are mainly replacement reactions;
i.e., cations such as Mn21 in slag is replaced by a solute in
hot metal (eg. 1/2 Si41) or the anion O22 is replaced by S
and two electrons from hot metal. The coke bed affects slag/
metal reactions by promoting the mixing of the two reacting
phases and/or lengthening the time available for reaction.
When the deadman coke bed is dirtied, the transit of all
liquids is prolonged and the opportunities for gas/metal,
coke/metal, and metal/slag reactions to proceed toward equilibrium are increased.
Let us start with an excellent hearth practice; i.e., the
deadman is very permeable and all liquids above the taphole
(for the simplicity of argument) are drained during each cast.
There are many causes which may lead to the beginning of
the deterioration of permeability of the deadman. For example, switching to weaker coke and/or less reducible ore;
increasing the generation of fines in the raceways; losing
the control of burden distribution; etc. The common consequences of these changes are the widening of the size distribution of coke and the decreasing of its average size. The
less permeable coke bed of the deadman has smaller and
fewer passages for liquid flows and exercises larger drag
force (due to larger solid-liquid contact area) to hinder these
flows. A drop in temperature in the hearth would cause
viscosities of liquids to increase and therefore hinder liquid
flows. With an increase in the production rate of liquids,
with the same casting schedule, the tapping rate has to
increase, which requires a more permeable deadman. Under
essentially the same operating conditions, the gradual deterioration of the permeability of an excellent deadman will
lead to slower flow rate of liquids inside the deadman, i.e.,
the hearth. Therefore, with the same casting practice, an
equivalent degree of liquid removal (dryness) can no
longer be achieved. Gradual increase in the amount of liquids
retained in the hearth means greater risk from excessive
liquid level in the case of any operating delay and/or higher
probability of wind volume (productivity) loss due to
decreased furnace permeability. Of course, if we know that
the permeability of the deadman is decreasing, then certain
countermeasures can be adopted.
Analysis techniques based on the extent of departure from
equilibrium states can yield valuable information on the state
of the coke bed in the furnace deadman and hearth. Previous
studies, most notably those of Tsuchiya et al.,[19] have related
departures from equilibrium for the partition of silicon, manganese, and sulfur between metal and slag phases to the
VOLUME 31B, OCTOBER 2000995

internal state of the furnace without making any reference

to the state or variation of the deadman coke bed. Their
effort appeared to be aimed at minimization of the fuel rate
rather than process monitoring and control as in the present
work. Tsuchiya et al. define partition reaction attainment to
0
equilibrium ratios, RSi 5 (LSi /L0Si)100, RMn 5 (LMn /LMn
)100,
and RS 5 (LS/LS0)100, where Li is the partitioning ratio calculated from the process data and L0i the equilibrium ratio. They
used the tapping slag composition and explained RSi behavior
in terms of the height of the furnace coke reserve zone and the
resultant reaction opportunity for silicon transfer to the hot metal
by the accepted primary mechanism involving SiO gas. This
explanation of thermal intensity is undoubtedly viable. Kinetic
considerations for heterogeneous reactions relating to the status
of the coke bed in the hearth were not included.
In the cases of RMn and RS, Tsuchiya et al. observed that the
kinetics of manganese and sulfur transfer reactions,
(Mn21) 1 2e [Mn] and [S] 1 2e (S22)
are more retarded when slag containing more residual FeO
enters the hearth to result in more of the following reaction:
(Fe21) 1 2e Fe
Under the conditions that raw materials quality was
essentially constant, they therefore relate these indexes to
burden descent stability. Again, the explanations made are
sound in fundamental terms and, in the case of sulfur, consistency with laboratory results is noted. However, the authors
did not discuss the possibility for variable extents of reaction
during dripping to influence manganese and sulfur transfer
directly or to influence the FeO load on the hearth.
If the permeability deteriorates from bad to worse, the
situation changes rather differently. The flow through the
deadman becomes too slow to be effective, particularly for
larger blast furnaces. Then, liquids from the active coke
zone have to find their own path of least resistance. An
increased amount of liquid will flow through the region
of high voidage, i.e., the gap between the impermeable
deadman core and the refractory sidewall. It is clearly more
risky to bring additional liquid iron (which is hot and undercarburized) and slag (which is oxidizing and contains alkali
oxides) in direct and close contact with carbon refractories.
D. The Carburization
The departure from equilibrium, which allows a direct
connection with the internal state of the furnace, is that for
carbon dissolution. This is because the information is carried
by the metal stream alone. This stream, while derived from
all the ferrous burden materials, is relatively uniform at the
time of discharge when compared to the often highly diverse
dripping slags. The high density and low viscosity of the
metal also minimize the opportunity for factors other than the
nature of the coke bed to influence kinetics of carburization.
Within the volume below the cohesive zone, a condition
of supply and demand exists for carbon. The saturation limit
is determined by the temperature of the hot metal and the
presence of other solute elements. The carbon to be supplied
is determined by the initial content of the dripping hot metal
(i.e., burden mix dependent), the reactivity of the carbon
sources present,[20] the available contact time, and the surface
area of contact.
996VOLUME 31B, OCTOBER 2000

In a blast furnace regime characterized by stable coke

quality, injectant load, burden mix, tuyere conditions, casting, and burden distribution practices, it is to be expected
that the supply/demand balance will be essentially maintained. Until recently,[21] it has been widely assumed that
the driving force and time available for carbon dissolution
are sufficient to ensure that saturation is the norm if not the
rule. In fact, hot metal tapped from the blast furnace is
almost always not saturated, and in general, the greater the
departure from saturation, the better the internal condition
and performance of the furnace. Carbon subsaturation of 0.2
pct or greater is observed for a clean deadman.
Once this nonsaturated state is recognized, the degree of
departure from equilibrium may be used to study the furnace
response to a change in any of the previously listed variables
determining carbon supply and demand.
In order to maintain high deadman permeability, it is
necessary that dripping metal consistently consumes fine
carbonaceous material. Of course, dripping metal contacts
survivor coke particles as well and dissolves carbon from
these too. It follows that when metal drains rapidly through
the clean deadman, it does little damage to the deadman
coke lumps and arrives in the hearth with a carbon appetite
able to promote hearth cleanliness and renewal with sound
coke. When the dripping metal is retained in a low permeability deadman, it causes damage to the deadman coke and
arrives in the hearth with little appetite to remove any fines.
The hearth coke bed renewal is retarded, and when it is
renewed, it is by previously damaged deadman coke. Therefore, once the deadman is dirtied, recovery is often very
difficult and prolonged.
Since messages carried by the composition of the metal
and slag streams relate to phenomena occurring in the whole
of the volume below the cohesive zone, it is only possible
to loosely ascribe locational variations in condition when
significant and sustained variations are observed between liquid streams tapped from individual tapholes. The only
other information available is that for the voidage condition
of the coke bed at the discrete level of the taphole level,
which can be estimated in continuous fashion by calculations
based on casting data as discussed elsewhere.[22]
V. DECODING MESSAGES CARRIED BY
LIQUID FLOWS AT PORT KEMBLA
Once it is understood that the status of the deadman coke
bed is implicated in the hot metal carbon content, other
messages about the behavior of dripping slags can be inferred
from the partitioning behavior of certain elements. These
behaviors also depend on the condition of the coke bed
through which the slag must also drip.
A. Metal Signals
The amount of carbon required for saturation is determined by the metal temperature and the concentration of
other elements also dissolved in the liquid iron. Interactions
between solute species occur at the atomic level and marked
changes in solute activities can arise. However, the linear
approximation of Eq. [1] has been found to be adequate.
At Port Kembla, an expression developed by Neumann
et al.[23] is used. This considers, that carbon saturation in
METALLURGICAL AND MATERIALS TRANSACTIONS B

iron containing small amounts of silicon, sulfur, phosphorous, and manganese is used to determine the carbon content
at saturation. The expression has the form
pct Csat 5 1.3 1 2.57 3 1023T
2 0.31 pct Si 2 0.33 pct P

[1]

2 0.4 pct S 1 0.028 pctMn

All concentrations are in mass percent and temperature
is in degrees Celsius. Although variations in the concentration of the other elements such as titanium must be acknowledged as potential sources of error, these have not proven
to have a debilitating effect on the application of this
expression.
The departure of hot metal carbon content from the saturation value, DC, is readily determined.
DC 5 pct Csat 2 pct Cactual

[2]

Because dirtying of the deadman by slag is also possible,

the hybrid deadman cleanliness index (DCI) has been developed by combining the temperature sensitive term of Eq. [1]
and the DC value with an expression describing sensitivity of
the liquidus temperature of the Port Kembla tapped slag to
its lime/silica basicity index.
1
DC
DCI 5 HMT 1
2.57 1023
2(1430 2 190 (1.23 2 C/S))

[3]

where HMT 5 hot metal temperature and C/S 5 the CaO/

SiO2 ratio of the tapped slag. The constant 2.57 1023 is
a conversion factor from the unit of concentration to the
unit of temperature; the liquidus temperature at C/S 5 1.23
is 1430 8C.
While there is little doubt that information on the primary
or dripping slags would be preferable, the previously expression has proven to be of greater utility in the Port Kembla
circumstance than the DC value alone.[21]
Dirtying of the deadman by raceway coke debris deposition can occur as a result of coke quality deterioration or
inappropriate increase in kinetic energy of the blast at the
tuyeres. Both conditions can be illustrated by operating data.
The former condition is illustrated with daily average (average of approximately 40 metal samples per day) in Figure
1, where coke of inferior quality was charged to the Port
Kembla No. 6 Blast Furnace for a period of several days
immediately prior to an extended maintenance shutdown on
February 22, 1997. The deterioration in DCI was immediate,
extreme, and prolonged with full recovery taking 12 months.
During this period, tuyere coke probe samples were taken
at the initial shutdown and on two subsequent occasions
(June 30 and October 28). Results for DCI, coke particle
size, and metal and slag retention at a depth of 2 m from
the tuyere nose are shown in Table I. Progressive improvements in hearth coke and liquid drainage at the standard
sampling time after shutdown are associated with increasing
values of DCI. These should be compared with the value
of 170 previously (and eventually later) experienced with
good quality coke and equivalent operations.
Once the deadman is dirtied by carbon fines in this way,
the removal can only occur by dissolution of the carbon in
the hot metal and reactions with slag.
METALLURGICAL AND MATERIALS TRANSACTIONS B

The influence of excessive blast kinetic energy is illustrated with daily average data in Figure 2, where operational
data for both Port Kembla furnaces are presented. In each
case, deadman cleanliness suffered immediately from the
time when increased tuyere velocity occurred as the result
of operating with small back-up blowing engines.
Dirtying of the deadman coke bed by freezing of slag or
unmelted ore is demonstrated in Figure 3, again using daily
average data. The DCI decreased sharply on the introduction
of titania bearing and unfluxed ore (52 pct Fe2O3, 33 pct
TiO2, 4 pct SiO2) to the No. 5 Blast Furnace in November
1988. This was charged to achieve a titania loading rate of
12 kg per tonne of hot metal in order to provide protection
to the hearth refractories. No other significant change to
furnace operations occurred in November, and it is clear
that the residue of the titania bearing ore, which may or may
not melt, impaired the permeability of the coke bed. It should
be noted that equivalent titania loading rates had been
charged via ilmenite/cement bricks (37 pct Fe2O3, 33 pct
TiO2, 5 pct SiO2, 7 pct Al2O3, 7 pct CaO) since February
1988 with little impact on DCI, and the dripping slag does
not appear to have suffered resolidification problems.
B. Slag/Metal Signals
Droplets of metal and slag must pass through the same
coke bed to reach the furnace hearth. Figure 4 shows daily
average data indicating that the partition ratios for manganese and titanium are highly correlated with DCI. Each
index data is presented in daily average form and involves
approximately 40 metal and 12 slag samples per day. The
behavior displayed is typical and is clearly consistent with
the proposal that all these indexes can be used to routinely
monitor changes to the permeability of the coke bed that is
common to the passage of both phases. A drop in DCI value
corresponds to a closer approach toward the equilibrium
condition for both carburization and the reduction of metal
oxides from slag. Changes occurring due to variations in the
coke quality and operational disruption may be monitored in
this way.
In 1984 to 1986, there were changes made to the iron
bearing raw materials used at Port Kembla. In this period,
the variations in titanium and manganese partitions were
rather different. Sinter has always supplied the bulk (,70
pct) of the ferrous burden, complemented by smaller quantities of pellets and lump ores. These materials are repeatedly
combined at the stockhouse to provide the same ferrous
layer at all levels of the furnace stack. Considerable mixing
of the materials within each layer also occurs on the charging
belt conveyor and during passage through the furnace charging elements.
In July 1985, a TiO2 and MgO bearing pellet was introduced into the furnace burden replacing lump ore. This pellet
replaced sinter and coke as the main sources of titania. Figure
5 shows daily average data for the blast furnace titania
loading, the hot metal temperature, the titania partition ratio,
and DCI for the period from June 1984 to December 1986.
High values of the Ti/TiO2 prior to the end of 1984 were
associated with higher concentrations of coke ash and the
presence of some titania minerals in the ash. Analysis records
for the ferrous burden streams were incomplete until 1985.
Fluctuations in Ti/TiO2 ratio during 1984 to 1985 are consistent with (inversely correlated to) DCI fluctuations. From
VOLUME 31B, OCTOBER 2000997

Fig. 1Extended recovery of Port Kembla No. 6 Blast Furnace from a period of poor coke quality coincident with an extended shutdown (circled).

Table I. Results of Tuyere Coke Probe Sampling No. 6

Blast Furnace
Shutdown Date
DCI
Pct-8-mm coke
Pct-4-mm coke
Mean coke size (mm)
Pct metallics
Pct slag

22/02/97

30/06/97

28/10/97

126
53
32
14
17
17

144
26
17
26
3
8

157
7
5
37
11
2

November 1985, the aim hot metal temperature was deliberately reduced and this was accompanied by decreases in
both DCI and Ti/TiO2. The former is to be anticipated from
Eq. [3], while the latter is consistent with the reduced thermodynamic driving force resulting from lower operating temperatures. Some increased variability of both the hot metal
temperature and the partition ratio is evident throughout
1986.
The utility of DCI as an indicator of overall furnace performance is demonstrated in Figure 6, where unedited data in
the form of monthly averages is presented for the entire
current campaign of Port Kembla No. 5 Blast Furnace (June
1991 to present). Hot metal silicon and, in particular, its
standard deviation provide the best available indicators of
thermal and process stability.
The significant correlation of the latter with DCI is particularly constructive. The fuel rate provides the most complete
available measure of process efficiency, while the total stave
heat load measures not only a significant component of
process energy loss but also a very major determinant for
campaign life.
998VOLUME 31B, OCTOBER 2000

C. Further Observations
It is also possible for the deadman to become dirtied by
an imbalance between carbon fines supply and demand due
to diminished carbon appetite of the hot metal. This can
occur in instances where aim hot metal temperature is
reduced or where the ferrous burden materials and operating
condition generate higher carbon contents in liquid iron at
meltdown.
Additionally, difficulties in maintaining deadman cleanliness can be intensified by increased metal concentrations of
those elements that decrease the saturation carbon content
of the metal, notably silicon, phosphorous and sulfur (Eq.
[1]). In the case of silicon, the consequence of the usage of
lump quartzite as a flux stone should be recognized, since
it has been shown that under Port Kembla circumstances,
some 43 pct of the silicon in lump quartzite reports directly
to the hot metal.[24] In the case of sulfur, very low concentrations (less than 0.010 pct) may also be problematic, since
the kinetics of carbon dissolution are significantly increased
for some materials in very low sulfur melts.[25] Under this
condition, the dripping metal may dissolve most carbon from
lump coke particles in the active coke zone and little in the
deadman. The average hot metal sulfur level at Port Kembla
is 0.014 pct.
In order to retain a healthy appetite for coke debris, char,
or graphite from the raceway, the lump coke itself should
have a low carburizing ability. This is quite opposed to recent
suggestions by Gudenau et al.[26] The inherent ability of the
coke to carburize the hot metal (e.g., the amount of ash and
its properties and distribution) can also impact on the ability
to keep the deadman clean.
During any blast furnace shut down, compaction of the
burden materials occurs after the upward force of the gas
flow is removed. Maintenance shutdowns at Port Kembla
METALLURGICAL AND MATERIALS TRANSACTIONS B

Fig. 2DCI responses to periods of increased tuyere velocity. Port Kembla Nos. 5 and 6 Blast Furnaces.

Fig. 3DCI response to the charging of titania ore. Port Kembla No. 5 Blast Furnace.

are typically of 36 or more hours duration. During this

time, a decrease of the burden stock line of 1 m or more is
commonly observed. This compaction accumulates across
the full height of the furnace contents. Recovery from compaction takes a much longer time in the volume of the
METALLURGICAL AND MATERIALS TRANSACTIONS B

deadman due to the weight of material above and the much

lower rate of material turnover in this zone. During extended
shutdowns, thermal losses also increase the probability of
retarded slag movement and even resolidification in post
startup operations. Following shutdowns, it is generally
VOLUME 31B, OCTOBER 2000999

Fig. 4Comparison between DCI and partition ratios for manganese and titanium. Port Kembla No. 5 Blast Furnace.

observed that the DCI and partition ratio deteriorations

extend for a period of several days beyond the full recovery
of hot metal temperature. This demonstrates the physical
nature of compromised permeability.
In addition to the ability to confirm DCI trends relating
to transient deadman condition, metalloid partition ratios
can occasionally be studied over the longer term to provide
some understanding about the behavior of primary or dripping slags. The composition of these slags often differs
greatly between raw material types. Where the input furnace
loading of any particular partitioning element is heavily and
consistently derived from a particular material, the partitioning ratio may provide useful insights into the behavior
of the dripping slag from that material.

1000VOLUME 31B, OCTOBER 2000

VI. DISCUSSION
In the chain of steps of processing raw materials to hot
metal, the weak link appears to be in the deadman-hearth
region. The sophistication and effectiveness in the control
of quality of raw materials, burden distribution, fuel injection, and tuyere practice reflect advances in our knowledge
about physical and chemical phenomena in those other
regions of the blast furnace. The present work is aimed to
shed some light on the regions below the cohesive zone
where there is no direct means to monitor the inner state.
The most critical part of this region is the coke bed through
which liquids flow.
The data presented here have indicated that there is significant information, which reflects the state of the deadman

METALLURGICAL AND MATERIALS TRANSACTIONS B

Fig. 5Responses of DCI and titanium partition ratio to decreased metal temperature. Port Kembla No. 5 Blast Furnace.

and hearth, carried out in tapped hot metal and slag. An

index, DCI, based on properties (i.e., daily average of temperature and chemical compositions) of hot metal and slag
is proposed as an indicator of the properties of the deadman
hearth. The most important property of the coke bed is liquid
permeability, which is sensitive to the average size and size
distribution of coke pieces and trapped oxide particles.
As a performance indicator, the usefulness of DCI has
been demonstrated. In Figure 6, particularly, the correlation
with silicon standard deviation indicates the relevance to
process stability. The partition of manganese and titanium
is shown in Figure 4.
Factors that cause DCI to change are shown in Figures
1, 2, 3, and 5. These factors determine the packing (properties
of solids) of the coke bed; i.e., they present a mechanism
for changes of permeability of the deadman. By accepting

METALLURGICAL AND MATERIALS TRANSACTIONS B

this mechanism, which is of course subject to further refinement when more data become available, operators know
what to expect at the time an adjustment is made.
The concept of a dynamic balance of the rate of generation
of fines in the lumpy zone and raceways and the rate of
consumption of fines in the deadman by hot metal and slag
is very important in the control of hearth condition. The
thermodynamic driving force and kinetic factors for the
flowing liquids to eliminate carbonaceous and oxide fines
are important and complex; however, they will not be a part
of this article. One must keep in mind the fact that the
operator does not have an effective tool (as the gas flow is
for the other part of the blast furnace) to correct conditions
in the hearth. Once the deadman is impaired, it takes a long
time, weeks and months, to recover, so the protection of the
deadman should have high priority.

VOLUME 31B, OCTOBER 20001001

at Port Kembla. Aim operating temperatures have been

increased. Tuyere diameters are being increased to reduce
blast kinetic energy. Coal blending and coal preparation
activities have been adjusted to significantly improve coke
performance in the furnace.
Current developmental activities also include laboratory
investigations to identify the expected carbon content of hot
metal droplets discharged from the cohesive zone for each
burden material in use. Mixtures of materials and sinters
representing development opportunities are also being examined. This is a simple and logical extension of existing
methods for studying the softening/melting behaviors of
these materials.
Research activities into coke quality are also now drawing
on DCI based learnings to explore new opportunities.
The lessons that can be derived from observation of DCI
and partition coefficients are extremely powerful because
of their fundamental basis. The signals are sometimes quite
noisy and require significant and appropriate filtering. Furthermore, these signals are inherently capable of reporting
on two or more concurrent process influences. For these
reasons, the task of message decoding is not straightforward
and the selection and filtering of data from suitable and
significant periods of operation continues to prove crucial.
As significant new data and behaviors are identified, they
are being systematically incorporated into a sophisticated
mathematical model of the process. These constitute
important improvements since the state of the hearth and
deadman have previously been the most crudely modeled
zones of the process. Sensitivity studies using the upgraded
model are now being used to identify areas for prioritized
developmental focus and to identify new questions that can
then be tested against the real process data sets. Continued
interplay and validation activities of this nature are expected
to generate advances in effective control of the hearth and
deadman, and ultimately in the whole of the ironmaking
process.

VIII. CONCLUSIONS

Fig. 6Comparison between DCI and key operating indices. Current campaign of Port Kembla No. 5 Blast Furnace.

Our understanding of hearth phenomena through the evaluation of DCI as presented in the present work is preliminary;
however, it provides the physical basis for the development
of mathematical models of hearth and casting practice. With
newly gained knowledge of hearth phenomena, the whole
blast furnace process becomes more or less transparent, i.e.,
without a blind spot. It might be the time to investigate and to
establish the required quality of raw materials and charging
practice for a given operating condition. The interplay
between modeling and observations will advance the technology of blast furnace ironmaking and lead to more efficient
commercial operations.
VII. FOCUS OF CURRENT EFFORTS
As a result of information derived from DCI based studies,
a number of process modifications have already been made
1002VOLUME 31B, OCTOBER 2000

The DCI developed at Port Kembla is based on the undersaturation of carbon in iron and the superheat of slag.
Experience with DCI has clearly shown that superior blast
furnace performance is associated with higher values of
this index.
By studying DCI responses to coke quality and operating
conditions, the critical importance of continually limiting
and consuming the quantity of carbonaceous debris from the
raceways and trapped oxide particles has been recognized.
Strategies to implement these learnings continue, and development activities to improve coke and ferrous material qualities are now being guided by DCI at Port Kembla.
Partition ratios of manganese and titanium are strongly
correlated to DCI and confirm that these fundamentally
based indices can also be used to continuously monitor the
condition of the lower zones of the furnace. Additionally,
where the total furnace load of a particular partitioning element is predominantly derived from a single raw material,
the behavior of the partition coefficient can allow conclusions to be drawn about the behavior of the dripping slag
derived from that material.
Observations from Port Kembla suggest that little mixing
METALLURGICAL AND MATERIALS TRANSACTIONS B

of primary slags occurs in the deadman and that most liquid

mixing occurs in the hearth pools.
ACKNOWLEDGMENTS
We are grateful to the Management of BHP Steel for
encouragement in this work and permission for its publication. The many constructive suggestions and conversations
of Port Kembla Ironmakers must also be acknowledged.
In particular, the statistical and data management support
provided by Kevin Price has been invaluable.
REFERENCES
1. J.G. Peacey and W.G. Davenport: The Iron Blast Furnace Theory and
Practice, Pergamon Press, Elmsford, NY, 1979.
2. M. Kondo, Y. Konishi, H. Koitabashi, Y. Morioka, S. Hasizume, H.
Takahashi, K. Okumura, and S. Tomita: Tetsu-to-Hagane, 1978, vol.
64, p. S546.
3. J.J. Bosley, N.B. Melchen, and M.M. Harris: J. Met., 1959, Sept., pp.
610-15.
4. V.M. Muravev and N.I. Mischenko: Stahl Eng., 1970, pp. 591-94.
5. Blast Furnace Phenomena and Modelling, Committee on Reaction
within Blast Furnace, Joint Society on Iron and Steel Basic Research,
The Iron and Steel Institute of Japan, Elsevier Applied Science, New
York, NY, 1987.
6. W.-K. Lu: 6th Int. Iron and Steel Congr., Nagoya, Oct. 2126, 1990,
The Iron and Steel Institute of Japan, Tokyo, Japan, 1990, vol. 2, pp.
548-57.
7. A. Poos: 6th Int. Iron and Steel Congr., Oct. 2126, 1990, Nagoya,
The Iron and Steel Institute of Japan, Tokyo, Japan, 1990, vol. 2, pp.
395-404.
8. A.K. Biswas: Principles of Blast Furnace Ironmaking, Cootha Pub.,
Brisbane, Australia, 1981, p. 139.
9. D. Lathlean, K. Edwards, S. Webb, and R. Nightingale: ICSTI/Ironmaking Conf. Proc., 1998, pp. 193-202.

METALLURGICAL AND MATERIALS TRANSACTIONS B

10. S. Wakuri, H. Kanoshima, M. Baba, T. Ashimura, M. Naito, and A.

Hatanaka: AIME Ironmaking Conf. Proc., 1989, pp. 573-86.
11. R.R. Willmers and R.M. Poultney: Cokemaking Int., 1992, vol. 4 (1),
pp. 69-78.
12. E. Beppler, B. Gerstenberg, U. Janhsen, and M. Peters: Requirements
on the Coke Properties Especially when Injecting High Coal Rates,
1992 Ironmaking Conference Proceedings, Toronto, Canada, pp. 17184.
13. K. Sasaki, M. Hatano, M. Watanabe, T. Shimoda, K. Yokotani, T. Ito,
and T. Yokoi: Tetsu-to-Hagane, 1976, vol. 62 (5), pp. 580-91.
14. K. Kambara, T. Hagiwara, A. Shigemi, S. Kondo, Y. Kanayama, K.
Wakabayashi, and Y. Hiramoto: Tetsu-to-Hagane, 1979, vol. 62 (5),
pp. 535-46.
15. K. Kojima, T. Nishi, T. Yamaguchi, H. Nakama, and S. Ida: Tetsu-toHagane, 1976, vol. 62 (5), pp. 570-79.
16. Y. Shimomura, Y. Kushima, and S. Arino: Report of 54th Committee
of Gakushin, Japan Society for Promotion of Science, Tokyo, Japan,
[Report No. 1484], 1979.
17. H. Takahashi, M. Tanno, and J. Katayama: Iron Steel Inst. Jpn. Int.,
1996, vol. 36 (1), pp. 1354-59.
18. W.-K. Lu: in Advances in Physical Chemistry of Process Metallurgy,
N. Sano, W.-K. Lu, and P. Riboud, Academic Press, London, 1997,
pp. 219-47.
19. N. Tsuchiya, S. Taguchi, Y. Takada, and K. Okabe: Tetsu-to-Hagane,
1977, vol. 63, pp. 1791-1800.
20. H.W. Gudenau, L. Meier, and V. Schemmann: ICSTI/Ironmaking Conf.
Proc., ISS, Toronto Canada, 1998, pp. 1068-72.
21. R.J. Nightingale, F.W.B.U. Tanzil, A.J.G. Beck, J.D. Dunning, and
S.K. Vardy: ICSTI/Ironmaking Conf. Proc., ISS, Toronto, Canada,
1998. pp. 567-72.
22. R.J. Nightingale and F.W.B.U. Tanzil: Ironmaker and Steelmaker,
1997, Feb., pp. 1-3.
23. F. Neumann, H. Schenck, and W. Patterson: Geierei-Tech. Wiss. Beihefte, 1959, vol. 23, pp. 1217-46.
24. K.P. Galvin, R.J. Nightingale, and A.G. Waters: AusIMM Proc., 1994,
vol. 299 (2), pp. 107-11.
25. B. McDonald, C. Wu, V. Sahajwall, K. Farrell, and T. Wall: ICST/
Ironmaking Conf. Proc., ISS, Toronto, Canada, 1998, pp. 1889-1900.
26. H.W. Gudenau, J.P. Mulanza, and D.G.R. Sharma: Steel Res., 1990,
vol. 61 (3), pp. 97-104.

VOLUME 31B, OCTOBER 20001003